Question 31
Domain 2: Core Machine Learning, AI, and Transformer FoundationsWhat is the purpose of teacher forcing in sequence generation?
Correct answer: B
Explanation
Teacher forcing trains a sequence model by feeding it the target output from the previous step instead of its own prediction. This is why it is described as using "ground truth tokens during training," which helps the model learn the correct next-token patterns more efficiently.
Why each option is right or wrong
A. To speed up inference
B. To use ground truth tokens during training
During training, sequence models are typically optimized by conditioning each next-step prediction on the actual previous target token rather than the model’s own earlier output, so the loss is computed against the true sequence at every timestep. This is a standard training technique in autoregressive models and is used to stabilize and speed up learning of next-token dependencies; it is not a legal or regulatory rule, so no statute or code section applies.
C. To reduce model size
D. To improve model interpretability